Docking Challenge: Protein Sampling and Molecular Docking

Article pubs.acs.org/jcim

Docking Challenge: Protein Sampling and Molecular Docking Performance Khaled M. Elokely† and Robert J. Doerksen*,†,‡ †

Department of Medicinal Chemistry, School of Pharmacy, University of Mississippi, University, Mississippi 38677, United States Research Institute of Pharmaceutical Sciences, School of Pharmacy, University of Mississippi, University, Mississippi 38677, United States

‡

S Supporting Information *

ABSTRACT: Computational tools are essential in the drug design process, especially in order to take advantage of the increasing numbers of solved X-ray and NMR protein−ligand structures. Nowadays, molecular docking methods are routinely used for prediction of protein−ligand interactions and to aid in selecting potent molecules as a part of virtual screening of large databases. The improvements and advances in computational capacity in the past decade have allowed for further developments in molecular docking algorithms to address more complicated aspects such as protein flexibility. The effects of incorporation of active site water molecules and implicit or explicit solvation of the binding site are other relevant issues to be addressed in the docking procedures. Using the right docking algorithm at the right stage of virtual screening is most important. We report a staged study to address the effects of various aspects of protein flexibility and inclusion of active site water molecules on docking effectiveness to retrieve (and to be able to predict) correct ligand poses and to rank docked ligands in relation to their biological activity for CHK1, ERK2, LpxC, and UPA. We generated multiple conformers for the ligand and compared different docking algorithms that use a variety of approaches to protein flexibility, including rigid receptor, soft receptor, flexible side chains, induced fit, and multiple structure algorithms. Docking accuracy varied from 1% to 84%, demonstrating that the choice of method is important.

1. INTRODUCTION Protein−ligand docking is a powerful tool to study and provide a proper understanding of protein−ligand interactions. Docking is regularly used in different stages of drug design strategies, such as to facilitate design of potentially active leads.1,2 Detection of the best ligand poses and proper ranking of several ligands’ relative docking propensity are of great importance. Molecular docking, in practice, has two essential requirements:3 structural data, for candidate ligands and the protein target of interest and a procedure to estimate protein− ligand interaction poses and strengths.4 The RSCB Protein Data Bank (PDB) repository5,6 is the main source of protein target structures for docking studies. The number of structures deposited in the PDB repository has been rapidly increasing for many years. Currently there are >62,000 PDB entries of protein−ligand complexes, of which >60,000 were solved by Xray and >1700 by NMR methods (other techniques were used to solve the remaining structures).7 The candidate ligands in docking procedures are generally small molecules. There was a rapid increase in the number of available synthesized chemical libraries after the development of combinatorial chemistry,8 which increased demand for the development of fast and cheap ways to test interactions with protein targets. The increasing numbers of PDB entries and of chemical database entries, coupled to the strong desire to be able to predict binding modes and binding affinities of ligands, has led to a wide © XXXX American Chemical Society

acceptance of the routine use of docking methods as a crucial step in virtual screening.9 Various molecular docking algorithms are available to predict protein−ligand poses and to rank them based on scoring functions implemented in each specific docking approach.10,11 Practically, docking software applications require protein−ligand sampling algorithms in order to be able to generate acceptable ligand poses. Ligand sampling algorithms, for ligand pose generation and placement in the active site, are of three types: shape matching,12,13 systematic search,14 and stochastic algorithms.15 Ligand conformational sampling is an essential step that generates a ligand multiconformer database to be used in ligand sampling. Conformational search is sometimes performed as a separate step before docking16 or can be implemented as an integrated part of the docking algorithm.17 Protein sampling refers to the allowed degree of binding site flexibility. Docking algorithms may consider the protein as a rigid body,12,18 as a soft body,14,19,20 to have flexible side chains,19−21 or to have certain flexible domains.22−24 Alternatively, protein flexibility can be represented by using multiple conformers or ensembles of rigid protein structures.25 Various classes of scoring functions are Special Issue: 2012 CSAR Benchmark Exercise Received: January 19, 2013

A

dx.doi.org/10.1021/ci400040d | J. Chem. Inf. Model. XXXX, XXX, XXX−XXX

Journal of Chemical Information and Modeling

Article

Figure 1. Study protocol starts with ligand and protein sampling, followed by setting up docking calculations with subsequent scoring, ranking, and pose prediction of the docked ligands. Protein sampling accounts for five approaches: rigid receptor, soft receptor, multiple (ensemble) receptors, FSC receptor, and IF receptor. Ligand sampling is either precedent to the docking procedure or a part of the specific docking technique. Scoring, ranking, and pose prediction are carried out in relation to known active and co-crystallized ligands.

used to estimate the binding affinities of ligand poses.26 Scoring functions can be classified as force field-based,27,28 empirical, 29,30 knowledge-based, 31,32 clustering and entropybased,33−35 or consensus scoring methods.36−38 Active site water molecules can be considered another aspect of docking target flexibility.39,40 Incorporation of active site water molecules in the docking procedure is challenging. Each water molecule needs to be analyzed to check if it is an integral part of the protein or just an artifact of the crystallization procedure. In this work, which was part of the annual Community Structure−Activity Resource (CSAR) challenge, we studied the ability of different protein−ligand molecular docking algorithms to regenerate the correct ligand binding mode of crystal structure bound ligands and to rank active ligands with respect to their activity data. The study involved a five-stage docking approach based on the degree of allowed protein flexibility (Figure 1).

4FSZ, 4FT0, 4GH2, 4FT3, 4FT5, 4FT7, 4FT9, 4FTA, and 4FTC). A database of 47 ligand structures was obtained from CSAR to be used in the primary study. Subsequently a final database of 184 ligands including the previous 47 was obtained from CSAR. 2.1.2. Extracellular Signal-Regulated Kinase 2 (ERK2). For primary study of ERK2 (also called mitogen-activated kinase 1 or MAPK1), we downloaded 3I5Z.pdb43 from the RCSB PDB repository. For secondary study, we downloaded 12 protein mol2 structures from CSAR and the same ones in pdb format from the RCSB PDB repository (PDB IDs: 4FUX, 4FUY, 4FV0, 4FV1, 4FV2, 4FV3, 4FV4, 4FV5, 4FV6, 4FV7, 4FV8, and 4FV9). The initial ligand database contained 39 structures, and the final database was extended to include a total of 52 structures. 2.1.3. UDP-3-O-N-Acetylglucosamine Deacetylase (LpxC) of Pseudomonas aeruginosa. The PDB structure of LpxC (PDB ID: 3P3E)44 was downloaded for the primary study. We downloaded five other protein structures in mol2 format from CSAR, four of which were deposited in pdb format. The PDB structures have been deposited in the RSCB PDB repository (PDB IDs: 4FW3, 4FW4, 4FW5, 4FW6, and 4FW7) but are not yet released. The initial ligand database consisted of 16 ligands, whereas the final database was extended to a total of 31 ligands. 2.1.4. Urokinase-Type Plasminogen Activator (UPA). We retrieved the pdb structure of UPA (also known simply as urokinase or urokinase plasminogen activator) from the PDB (PDB ID: 1OWE).45 Seven UPA structures were downloaded from CSAR as mol2 files and from the RSCB PDB repository as pdb files (PDB IDs: 4FU7, 4FU8, 4FU9, 4FUB, 4FUC, 4FUD, and 4FUE). The initial ligand database consisted of 20

2. METHODS 2.1. Ligand Databases Collection. Ligand databases for all target proteins in the study were downloaded from CSAR. Initial databases did not include the activity data, whereas the final databases did include it. Protein-bound ligands were included in both the initial and final databases to serve as a check on pose prediction accuracy. 2.1.1. Serine/Threonine Protein Kinase Chk1 (CHK1). The structure of CHK1 (also called checkpoint kinase 1) was downloaded from the RCSB PDB repository (PDB ID: 2E9N)41 and used for primary study. A database of 17 additional PDB structures was downloaded in mol2 format from CSAR42 and in pdb format from the RCSB PDB (PDB IDs: 4FSM, 4FSN, 4FSQ, 4FSR, 4FST, 4FSU, 4FSW, 4FSY, B



Article

structures, but that was extended to 46 structures in the final database. 2.2. Protein Preparation with Protein Preparation Wizard.46 The PDB protein−ligand structures were processed with the Protein Preparation Wizard in the Schrödinger suite.47 The protein structure integrity was checked and adjusted, and missing residues and loop segments near the active site were added using Prime.48−50 Hydrogen atoms were added after deleting any original ones, followed by adjustment of bond orders for amino acid residues and the ligand. The protonation and tautomeric states of Asp, Glu, Arg, Lys, and His were adjusted to match a pH of 7.4. Possible orientations of Asn and Gln residues were generated. Active site water molecules beyond 5.0 Å from the ligand were deleted. Hydrogen bond sampling with adjustment of active site water molecule orientations was performed using PROPKA51 (propka.ki.ku.dk) at pH 7.4. Water molecules with fewer than two hydrogen bonds to non-waters were deleted. Then, the protein−ligand complex was subjected to geometry refinement using an OPLS2005 force field52 restrained minimization with convergence of heavy atoms to an RMSD of 0.3 Å. 2.3. Ligand Preparation with Ligprep.53 Ligands were prepared using Ligprep from the Schrödinger suite. We obtained the initial ligand databases as collections of SMILES (simplified molecular-input line-entry system) strings (which do not contain 3D coordinates). The final ligand databases were in the mol2 format (3D structures). We included all structures without performing predocking filtering. We generated a single low energy 3D conformer with acceptable bond lengths and angles for each 2D structure in the initial databases. For the initial and final databases, after 3D structure generation, we prepared ligand structures for molecular docking. Ligprep used the OPLS2005 force field and charges in all ligand preparation steps. All possible protomers (protonation states) and ionization states were enumerated for each ligand using Ionizer at a pH of 7.4. Stereoisomers were generated for the five structures with unassigned stereogenic centers, with a limit of 32 stereoisomers considered per ligand. Tautomeric states were generated for chemical groups with possible prototropic tautomerism. Only the lowest energy conformer was kept for each ligand. 2.4. Ligand Conformational Sampling. Ligand conformational sampling was used with the FRED54 and HYBRID55 modules of OpenEye. Other molecular docking software applications we used in this study have their own integrated conformational sampling algorithms. We used OMEGA 2.4.6 (OpenEye)56,57 as the conformer generator. The SMILES notations of the initial databases and mol2 structures of the final databases were used as input for OMEGA 2.4.6. The conformational search force field was defined as the 94s variant of the Merck Molecular Force Field (MMFF94s).58 We kept all generated conformers within a 10.0 kcal/mol energy window, except that to eliminate redundant conformers an RMSD cutoff of 0.5 Å was used. 2.5. Protein Sampling. To address protein flexibility in molecular docking, a large number of degrees of freedom should be considered. We performed a staged study starting from rigid body docking and continuing all the way to fully flexible active site docking. 2.5.1. Docking Using Rigid Receptors. The OEDocking v3.0.0 distribution using the FRED ligand shape fitting algorithm was utilized for receptor rigid body docking. All receptors used in this study were co-crystallized with ligands.

The bound ligands were used to specify the active site. A 3D box was generated around each ligand to enclose the active site. Because we do not have any extremely large active site in our study, a negative image potential was created for each active site with disabled inner contour. No constraints were added except for the LpxC target, for which we prepared receptor docking sites with and without a Zn2+ metal constraint. We prepared each receptor with and without active site water molecules. We saved the multiconformer ligand files in OEBinary, and therefore, there was no need to use the FRED conformer test flag. FRED was used with standard docking precision using 1.0 Å for the ligand translational step size and 1.5 Å for the rotational step size. Because the databases we used are small, we maintained the default value of keeping the 500 top scoring molecules with a maximum of one pose to be saved for each molecule. 2.5.2. Docking Using Soft Receptors. We used Glide 5.859 for soft receptor molecular docking. The receptor grid for each target was prepared using the OPLS2005 force field. We specified the area surrounding the co-crystallized ligand as the receptor binding pocket. We excluded all bound ligands from the grid generation, except for Zn2+ in the case of LpxC and active site water molecules when applicable. Softening the potential of the nonpolar parts of the receptor was carried out by scaling the van der Waals radii by a factor of 0.8. Atoms were considered as nonpolar if their absolute partial atomic charge was determined to be ≤0.25. The grid center was set to be the centroid of the co-crystallized ligand, and the cubic grid had a side length of 10 Å. No constraints were used in any of the receptor grids. Rotations were allowed for the hydroxyl groups in Ser, Thr, and Tyr, and the thiol group in Cys residues. After grid preparation, prepared ligand databases were docked into the generated receptor grids using Glide SP docking precision. Flexible ligand sampling was considered in the docking procedure. For the docking runs, a second softening potential was considered. A 0.8 scaling factor was used for van der Waals radii of the ligands’ nonpolar atoms with absolute partial atomic charge ≤0.15. All poses were subjected to post-docking minimization. We saved the best-docked structure for each ligand, based on the model energy score which combines the energy grid score, the binding affinity, the internal strain energy, and the Coulomb−van der Waals interaction energy scores. 2.5.3. Docking Using Receptors with Flexible Side Chains. We used the protein−ligand ant system (PLANTS)60 to deal with side chain flexibility of amino acid residues. PLANTS uses the artificial ant colony optimization (ACO) algorithm to find the best ligand pose in the binding pocket. ZODIAC 0.6561 was used to prepare PLANTS input files. All ligands and protein structures were preprocessed by the structure protonation and recognition system (SPORES)62 to adjust the protonation and tautomeric states and to assign stereoisomers for non-specified asymmetric centers. The binding site was specified from the cocrystallized ligand coordinates. We used normal ligand sampling search speed with 20 ants and simplex rescoring. We used values for the sigma parameter (σ = 1.0) and evaporation rate (ρ = 0.5) that have been shown to be sufficient with 20 ants.63 Planar bond rotations were forced. For clustering, an RMSD of 2 Å and a maximum of 10 clusters were defined. CHEMPLP64 was specified as the scoring function. All amino acids in the defined binding site were selected to have flexible side chains during docking. 2.5.4. Docking Using Flexible Binding Domain for Receptors, or Induced Fit Docking (IFD). Ligands are known C



Article

Figure 2. Active site structures of target proteins; CHK1 (PDB ID: 2E9N, panel A), ERK2 (PDB ID: 3I5Z, panel B), LpxC (PDB ID: 3P3E, panel C) and UPA (PDB ID: 1OWE, panel D). The images were generated using PyMol.72 The α-helices, β-sheets, and loops are colored in yellow, blue, and green, respectively. Key amino acids in the active sites have their side chains displayed as lines and are labeled based on their position. Only polar hydrogen atoms are displayed in white. All Cα atoms are represented as spheres and colored according to the corresponding secondary structure. Other carbons of the amino acids are colored green. Ligands are displayed as sticks with gray carbons, and no hydrogen atoms are depicted. Oxygen, nitrogen, and sulfur atoms are colored red, blue, and yellow, respectively.

The Schrödinger suite has the same capability of ensemble docking but with a soft receptor approach. Receptor grids were prepared as described above. The van der Waals scaling factor was specified as 0.8 for receptor nonpolar atoms, and a partial charge cutoff of 0.15 was used. Glide SP was used for docking, and one pose was saved for each ligand. 2.6. Pose Prediction of Single Ligands and Pose Fitting of Native Ligands. For more accurate pose fitting and prediction, we tried POSIT v.1.0.2 from the OpenEye suite.65 We prepared the receptors and allowed mild ligand−protein clashes in the generated receptors to account for the average coordinate error expected in PDB structures. We used the combine-receptor option to allow for identifying and subsequent use of pockets that would be unexplored if we used only single PDB files. The merged receptors were used in the FRED docking step as well. We allowed alternate posing of each ligand within 0.5 Å RMSD in each receptor. Mild clashes similar to those used in receptor preparation were allowed during pose prediction. We forced aromatic rings to be planar. The minimum probability to accept poses within 2.0 Å of the native ligand was set to 0.33 with minimum initial probability of 0.05. Receptors that had initial rigid TanimotoCombo < 0.866 were rejected. All generated protein−ligand complexes were subjected to a final optimization preserving the interactions

often to induce conformational changes in the active site upon binding. We used the Schrödinger induced fit docking (IFD) protocol to represent this. The receptor grid center was specified from the bound ligand, and the cubic grid had a side length of 10 Å. A 2.5 kcal/mol energy window that was used for ligand conformational sampling. The scaling factors to soften the potentials of the receptors and ligands were set to 0.5 in both cases. A maximum of 20 poses was saved. All residues within 5.0 Å of ligand poses were refined using the Prime molecular dynamics module to allow for binding domain flexibility. The Schrödinger IFD software allows specification of whole loops to move, but we found that if the key linking residues are flexible then the whole loop can adjust to fold differently without needing to use that option. Glide SP was used for the redocking step into the top 20 receptor structures generated within 30 kcal/mol of the best structure by the Prime refinement. 2.5.5. Docking Using an Ensemble of Protein Structures. Another way to address protein flexibility is to use multiple or an ensemble of protein structures. OEDocking v3.0.0 with its HYBRID program was used for this. We used the same options as in FRED docking but with multiple structures of each target. All protein structures were treated as rigid. D



Article

Figure 3. Structure of CHK1 (PDB ID: 2E9N, panel A) is displayed as blue cartoon. The most flexible residues have side chains represented as red lines (inside the yellow box). This flexibility is inferred from aligning the CHK1 active site from multiple PDB structures (panel B). The α-helices, βsheets, and loops in panel B are yellow, blue, and green, respectively. Images were generated using PyMol.

Figure 4. Structure of ERK2 (PDB ID: 3I5Z, panel A) is displayed as blue cartoon. The most flexible residues have side chains represented as red lines (inside the yellow box). The flexibility of the active site is shown by the alignment of the ERK2 binding site as determined from multiple PDB structures (panel B). The α-helices, β-sheets, and loops in panel B are yellow, blue, and green, respectively. Images were generated using PyMol.

the correct preparation of the target protein structure.67 All PDB structures that were used in this study were checked for integrity, especially of the active site region, and any structure which showed gap(s) in the binding site region was rejected. Some reports advocate not to include any geometry refinement step prior to generating the receptor for docking because this may incorrectly improve the protein−ligand interactions.68 Protein−ligand complexes were minimized before preparing the receptor grid for use in the Glide modules. This minimization step was not implemented for the other software applications. The active site was defined by the amino acid residues surrounding the bound ligand (Figure 2). Protein conformational changes often take place upon ligand binding, so ignoring protein flexibility during molecular docking may give results that are incorrect.69 There are several approaches to include protein flexibility in the docking procedure. We tried rigid body docking with FRED, soft receptor docking by softening the potential of receptor nonpolar regions with Glide, docking using receptors with flexible side chains with PLANTS, docking in multiple protein structures with HYBRID and Glide ensemble docking, and IFD. Each technique has its own approximations, advantages, and limitations.

associated with atoms involved in the TanimotoCombo score. A cutoff of 10 kcal/mol was used as the maximum strain to accept.

3. RESULTS AND DISCUSSION Ligand recognition by a protein depends on both shape (3D structure) and electrostatic complementarities. Several reports have described the effect of ligand and protein preparation steps on molecular docking efficiency. We used LigPrep, Protein Preparation Wizard, and SPORES to adjust the protonation, tautomeric, and stereoisomeric states of protein and ligand databases. Ligand conformational sampling is as important as correct ligand preparation. FRED and HYBRID require multiconformer databases to be prepared separately. Glide modulesGlide docking, Glide ensemble docking, and Glide induced-fit docking (IFD)PLANTS and POSIT each have an integrated conformational search algorithm. Pregeneration of conformers could be more beneficial if we have ligand databases with saturated rings, and non-specified E/Z vinyl, E/ Z amide bonds, and stereogenic centers, and if the integrated conformational search algorithm has limited options to handle these situations. Molecular docking efficiency is influenced by E



Article

Figure 5. Structure of LpxC (PDB ID: 3P3E, panel A) is displayed as blue cartoon. The most flexible residues have side chains represented as red lines (inside the yellow box). The flexibility of the active site is shown by the alignment of the LpxC binding site as determined from multiple PDB structures (panel B). The α-helices, β-sheets, and loops in panel B are yellow, blue, and green, respectively. Images were generated using PyMol.

Figure 6. Structure of UPA (PDB ID: 1OWE, panel A) is displayed as blue cartoon. The most flexible residues have side chains represented as red lines (inside the yellow box). The flexibility of the active site is shown by the alignment of the UPA binding site as determined from multiple PDB structures (panel B). The α-helices, β-sheets, and loops in panel B are yellow, blue, and green, respectively. Images were generated using PyMol.

Figure 7. Binding area of ERK2 as defined by a single PDB structure, 3I5Z, is shown in orange surface mesh (around the ligand) and dark green surface mesh (extended to the protein surface close to the ligand) (left), and the merged area as defined by two PDB structures, 3I5Z and 4FUX, as shown in light green surface mesh surrounding the ligand and extended protein surface from 3I5Z alone (right). The α-helices, β-sheets, and loops are red, yellow, and white, respectively. Images were generated using VIDA.73.

As a first step, we aligned the binding sites of the PDB structures of each target to check for conformational flexibility of active site residues. The four targets have different degrees of

flexibility. CHK1 showed a high degree of flexibility in the Ploop region (residues 13−23) and for the side chains of Lys38, Glu55, Val68, Lys91, Ser147, and Asp148 (Figure 3). The PF



Article

Table 1. Docking Algorithm Features, Computational Times, and Percent Accuracies ensemble docking scoring function conformational sampling comp. time/ligand (min) protein CHK1 ERK2 LpxC UPA a

rigid

soft

Chemgauss4 pregeneration 70% to bind within the active site of the proteins. Greater than 90% of the inactive ligands did not show any binding probability. The scoring functions are another important aspect to address because they play an important role in ligand ranking and pose selection. In OEDocking, the scoring function is used to select the best pose and ligand placement in the active site is based on a shape-fitting algorithm. In Glide and flexible side chain algorithms, ligand posing and ranking are based solely on the scoring function. We found that OEDocking performed the best. Utilization of the newly implemented chemical Gaussian overlay (CGO) function71 may enhance OEDocking performance even more.

4. CONCLUSION The best docking technique should be chosen after studying in detail the target, candidate ligands, and docking method performance. Benchmark analysis should be considered before J



■

Article

(14) Friesner, R. A.; Banks, J. L.; Murphy, R. B.; Halgren, T. A.; Klicic, J. J.; Mainz, D. T.; Repasky, M. P.; Knoll, E. H.; Shelley, M.; Perry, J. K.; Shaw, D. E.; Francis, P.; Shenkin, P. S. Glide: A new approach for rapid, accurate docking and scoring. 1. Method and assessment of docking accuracy. J. Med. Chem. 2004, 47 (7), 1739− 1749. (15) Thomsen, R.; Christensen, M. H. MolDock: A new technique for high-accuracy molecular docking. J. Med. Chem. 2006, 49 (11), 3315−3321. (16) McGann, M. FRED pose prediction and virtual screening accuracy. J. Chem. Inf. Model. 2011, 51 (3), 578−596. (17) Neves, M.; Totrov, M.; Abagyan, R. Docking and scoring with ICM: The benchmarking results and strategies for improvement. J. Comput.-Aided Mol. Des. 2012, 26 (6), 675−686. (18) McGaughey, G. B.; Sheridan, R. P.; Bayly, C. I.; Culberson, J. C.; Kreatsoulas, C.; Lindsley, S.; Maiorov, V.; Truchon, J.-F.; Cornell, W. D. Comparison of topological, shape, and docking methods in virtual screening. J. Chem. Inf. Model. 2007, 47 (4), 1504−1519. (19) Jones, G.; Willett, P.; Glen, R. C. Molecular recognition of receptor sites using a genetic algorithm with a description of desolvation. J. Mol. Biol. 1995, 245 (1), 43−53. (20) Jones, G.; Willett, P.; Glen, R. C.; Leach, A. R.; Taylor, R. Development and validation of a genetic algorithm for flexible docking. J. Mol. Biol. 1997, 267 (3), 727−748. (21) Korb, O.; Cole, J., Ant Colony Optimization for Ligand Docking. In Proceedings of the 7th International Conference on Swarm Intelligence; Springer-Verlag: Brussels, Belgium, 2010; pp 72−83. (22) Cavasotto, C. N.; Abagyan, R. A. Protein flexibility in ligand docking and virtual screening to protein kinases. J. Mol. Biol. 2004, 337 (1), 209−225. (23) Barreca, M. L.; Iraci, N.; De Luca, L.; Chimirri, A. Induced-fit docking approach provides insight into the binding mode and mechanism of action of HIV-1 integrase inhibitors. ChemMedChem 2009, 4 (9), 1446−1456. (24) Davis, I. W.; Baker, D. Rosetta ligand docking with full ligand and receptor flexibility. J. Mol. Biol. 2009, 385 (2), 381−392. (25) Huang, S.-Y.; Zou, X. Ensemble docking of multiple protein structures: Considering protein structural variations in molecular docking. Proteins: Struct., Funct., Bioinf. 2007, 66 (2), 399−421. (26) Huang, S.-Y.; Grinter, S. Z.; Zou, X. Scoring functions and their evaluation methods for protein−ligand docking: recent advances and future directions. Phys. Chem. Chem. Phys. 2010, 12 (40), 12899− 12908. (27) Yin, S.; Biedermannova, L.; Vondrasek, J.; Dokholyan, N. V. MedusaScore: An accurate force field-based scoring function for virtual drug screening. J. Chem. Inf. Model. 2008, 48 (8), 1656−1662. (28) Huang, N.; Kalyanaraman, C.; Irwin, J. J.; Jacobson, M. P. Physics-based scoring of protein−ligand complexes: Enrichment of known inhibitors in large-scale virtual screening. J. Chem. Inf. Model. 2005, 46 (1), 243−253. (29) Böhm, H.-J. The development of a simple empirical scoring function to estimate the binding constant for a protein−ligand complex of known three-dimensional structure. J. Comput.-Aided Mol. Des. 1994, 8 (3), 243−256. (30) Eldridge, M. D.; Murray, C. W.; Auton, T. R.; Paolini, G. V.; Mee, R. P. Empirical scoring functions: I. The development of a fast empirical scoring function to estimate the binding affinity of ligands in receptor complexes. J. Comput.-Aided Mol. Des. 1997, 11 (5), 425−445. (31) Muegge, I.; Martin, Y. C. A general and fast scoring function for protein−ligand interactions: A simplified potential approach. J. Med. Chem. 1999, 42 (5), 791−804. (32) Gohlke, H.; Hendlich, M.; Klebe, G. Knowledge-based scoring function to predict protein−ligand interactions. J. Mol. Biol. 2000, 295 (2), 337−356. (33) Ruvinsky, A. Calculations of protein-ligand binding entropy of relative and overall molecular motions. J. Comput.-Aided Mol. Des. 2007, 21 (7), 361−370. (34) Chang, M. W.; Belew, R. K.; Carroll, K. S.; Olson, A. J.; Goodsell, D. S. Empirical entropic contributions in computational

ASSOCIATED CONTENT

S Supporting Information *

Docking scores (y-axis) versus reported experimental activity data for each of the four targets and six docking algorithms. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Phone: 662-915-5880. Fax: 662-9155638. Address: 21 Faser Hall, Department of Medicinal Chemistry, University of Mississippi, University, Mississippi, 38677-1848. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS Thanks to Ronak Y. Patel for helpful discussions. This work was supported in part by the National Science Foundation under Awards EPS 0903787 and EPS 1006883. This investigation was conducted in part in a facility constructed with support from research facilities improvement program C06 RR-14503-01 from the NIH NCRR.

■

REFERENCES

(1) Huang, S.-Y.; Zou, X. Advances and challenges in protein−ligand docking. Int. J. Mol. Sci. 2010, 11 (8), 3016−3034. (2) Sousa, S. F.; Fernandes, P. A.; Ramos, M. J. Protein−ligand docking: Current status and future challenges. Proteins: Struct., Funct., Bioinf. 2006, 65 (1), 15−26. (3) Novikov, F. N.; Chilov, G. G. Molecular docking: Theoretical background, practical applications and perspectives. Mendeleev Commun. 2009, 19 (5), 237−242. (4) Balaji, G. A.; Balaji, V. N.; Rao, S. N. Utility of scoring function customization in docking-based virtual screening approaches. Curr. Sci. 2013, 104 (1), 86−97. (5) Berman, H. M.; Westbrook, J.; Feng, Z.; Gilliland, G.; Bhat, T. N.; Weissig, H.; Shindyalov, I. N.; Bourne, P. E. The Protein Data Bank. Nucleic Acids Res. 2000, 28, 235−242. (6) Bernstein, F. C.; Koetzle, T. F.; Williams, G. J. B.; Meyer, E. F., Jr; Brice, M. D.; Rodgers, J. R.; Kennard, O.; Shimanouchi, T.; Tasumi, M. The protein data bank: A computer-based archival file for macromolecular structures. Arch. Biochem. Biophys. 1978, 185 (2), 584−591. (7) RCSB Protein Data Bank (PDB). http://www.rcsb.org/ (accessed November 16, 2012). (8) Mario Geysen, H.; Schoenen, F.; Wagner, D.; Wagner, R. Combinatorial compound libraries for drug discovery: An ongoing challenge. Nat. Rev. Drug Discovery 2003, 2 (3), 222−230. (9) Scior, T.; Bender, A.; Tresadern, G.; Medina-Franco, J. L.; Martínez-Mayorga, K.; Langer, T.; Cuanalo-Contreras, K.; Agrafiotis, D. K. Recognizing pitfalls in virtual screening: A critical review. J. Chem. Inf. Model. 2012, 52 (4), 867−881. (10) Onodera, K.; Satou, K.; Hirota, H. Evaluations of molecular docking programs for virtual screening. J. Chem. Inf. Model. 2007, 47 (4), 1609−1618. (11) Cross, J. B.; Thompson, D. C.; Rai, B. K.; Baber, J. C.; Fan, K. Y.; Hu, Y.; Humblet, C. Comparison of several molecular docking programs: Pose prediction and virtual screening accuracy. J. Chem. Inf. Model. 2009, 49 (6), 1455−1474. (12) McGann, M. R.; Almond, H. R.; Nicholls, A.; Grant, J. A.; Brown, F. K. Gaussian docking functions. Biopolymers 2003, 68 (1), 76−90. (13) Jain, A. N. Surflex: Fully automatic flexible molecular docking using a molecular similarity-based search engine. J. Med. Chem. 2003, 46 (4), 499−511. K



Article

docking: Evaluation in APS reductase complexes. J. Comput. Chem. 2008, 29 (11), 1753−1761. (35) Lee, J.; Seok, C. A statistical rescoring scheme for protein− ligand docking: Consideration of entropic effect. Proteins: Struct., Funct., Bioinf. 2008, 70 (3), 1074−1083. (36) Liu, S.; Fu, R.; Zhou, L.-H.; Chen, S.-P. Application of consensus scoring and principal component analysis for virtual screening against β-secretase (BACE-1). PLoS One 2012, 7 (6), e38086. (37) Feher, M. Consensus scoring for protein−ligand interactions. Drug Discovery Today 2006, 11 (9−10), 421−428. (38) Teramoto, R.; Fukunishi, H. Consensus scoring with feature selection for structure-based virtual screening. J. Chem. Inf. Model. 2008, 48 (2), 288−295. (39) Ross, G. A.; Morris, G. M.; Biggin, P. C. Rapid and accurate prediction and scoring of water molecules in protein binding sites. PLoS One 2012, 7 (3), e32036. (40) Corbeil, C. R.; Moitessier, N. Docking ligands into flexible and solvated macromolecules. 3. Impact of input ligand conformation, protein flexibility, and water molecules on the accuracy of docking programs. J. Chem. Inf. Model. 2009, 49 (4), 997−1009. (41) Tong, Y.; Claiborne, A.; Stewart, K. D.; Park, C.; Kovar, P.; Chen, Z.; Credo, R. B.; Gu, W.-Z.; Gwaltney, S. L., II; Judge, R. A.; Zhang, H.; Rosenberg, S. H.; Sham, H. L.; Sowin, T. J.; Lin, N.-H. Discovery of 1,4-dihydroindeno[1,2-c]pyrazoles as a novel class of potent and selective checkpoint kinase 1 inhibitors. Bioorg. Med. Chem. 2007, 15 (7), 2759−2767. (42) Community Structure−activity Resource (CSAR). http://www. csardock.org/ (accessed April 2, 2013). (43) Aronov, A. M.; Tang, Q.; Martinez-Botella, G.; Bemis, G. W.; Cao, J.; Chen, G.; Ewing, N. P.; Ford, P. J.; Germann, U. A.; Green, J.; Hale, M. R.; Jacobs, M.; Janetka, J. W.; Maltais, F.; Markland, W.; Namchuk, M. N.; Nanthakumar, S.; Poondru, S.; Straub, J.; ter Haar, E.; Xie, X. Structure-guided design of potent and selective pyrimidylpyrrole inhibitors of extracellular signal-regulated kinase (ERK) using conformational control. J. Med. Chem. 2009, 52 (20), 6362−6368. (44) Lee, C.-J.; Liang, X.; Chen, X.; Zeng, D.; Joo, S. H.; Chung, H. S.; Barb, A. W.; Swanson, S. M.; Nicholas, R. A.; Li, Y.; Toone, E. J.; Raetz, C. R. H.; Zhou, P. Species-specific and inhibitor-dependent conformations of LpxC: Implications for antibiotic design. Chem. Biol. 2011, 18 (1), 38−47. (45) Wendt, M. D.; Rockway, T. W.; Geyer, A.; McClellan, W.; Weitzberg, M.; Zhao, X.; Mantei, R.; Nienaber, V. L.; Stewart, K.; Klinghofer, V.; Giranda, V. L. Identification of novel binding interactions in the development of potent, selective 2-naphthamidine inhibitors of urokinase. Synthesis, structural analysis, and SAR of Nphenyl amide 6-substitution. J. Med. Chem. 2003, 47 (2), 303−324. (46) Schrödinger Suite 2012: Protein Preparation Wizard; Schrödinger, LLC: New York. (47) Schrödinger Suite 2012; Schrödinger, LLC: New York. (48) Jacobson, M. P.; Pincus, D. L.; Rapp, C. S.; Day, T. J. F.; Honig, B.; Shaw, D. E.; Friesner, R. A. A hierarchical approach to all-atom protein loop prediction. Proteins: Struct., Funct., Bioinf. 2004, 55 (2), 351−367. (49) Jacobson, M. P.; Friesner, R. A.; Xiang, Z.; Honig, B. On the role of the crystal environment in determining protein side-chain conformations. J. Mol. Biol. 2002, 320 (3), 597−608. (50) Schrödinger Suite 2012: Prime, version 3.1; Schrödinger, LLC: New York. (51) Rostkowski, M.; Olsson, M.; Sondergaard, C.; Jensen, J. Graphical analysis of pH-dependent properties of proteins predicted using PROPKA. BMC Structural Biology 2011, 11 (1), 6. (52) Jorgensen, W. L.; Tirado-Rives, J. Potential energy functions for atomic-level simulations of water and organic and biomolecular systems. Proc. Natl. Acad. Sci. U.S.A. 2005, 102 (19), 6665−6670. (53) Schrödinger Suite: LigPrep, version 2.5; Schrödinger, LLC: New York, 2012.

(54) OEDocking v3.0.0: FRED v3.0.0; OpenEye Scientific Software: Santa Fe, NM. http://www.eyesopen.com (accessed April 2, 2013). (55) OEDocking v3.0.0: HYBRID v3.0.0; OpenEye Scientific Software: Santa Fe, NM. http://www.eyesopen.com (accessed April 2, 2013). (56) OMEGA v2.4.6; OpenEye Scientific Software: Santa Fe, NM. http://www.eyesopen.com (accessed April 2, 2013). (57) Hawkins, P. C. D.; Skillman, A. G.; Warren, G. L.; Ellingson, B. A.; Stahl, M. T. Conformer generation with OMEGA: Algorithm and validation using high quality structures from the protein databank and cambridge structural database. J. Chem. Inf. Model. 2010, 50 (4), 572− 584. (58) Halgren, T. A. MMFF VI. MMFF94s option for energy minimization studies. J. Comput. Chem. 1999, 20 (7), 720−729. (59) Schrödinger Suite 2012: Glide, version 5.8; Schrödinger, LLC: New York, 2012. (60) Korb, O.; Stützle, T.; Exner, T. E. Protein−Ligand ANT System, version 1.2; Universität Konstanz, Fachbereich Chemie und Zukunftskolleg: Konstanz, Germany, 2012. (61) Zonta, N. ZODIAC, version 0.65 beta. http://www.zeden.org (accessed April 2, 2013). (62) ten Brink, T.; Exner, T. E. Structure PrOtonation and REcognition System (SPORES), version 1.28; Universität Konstanz, Fachbereich Chemie und Zukunftskolleg: Konstanz, Germany, 2012. (63) Korb, O.; Stüzle, T.; Exner, T., PLANTS: Application of Ant Colony Optimization to Structure-Based Drug Design. In Ant Colony Optimization and Swarm Intelligence; Dorigo, M., Gambardella, L., Birattari, M., Martinoli, A., Poli, R., Stützle, T., Eds.; Springer: Berlin, Heidelberg, 2006; Vol. 4150, pp 247−258. (64) Korb, O.; Stützle, T.; Exner, T. E. Empirical scoring functions for advanced protein−ligand docking with PLANTS. J. Chem. Inf. Model. 2009, 49 (1), 84−96. (65) POSIT, version 1.0.2; OpenEye Scientific Software: Santa Fe, NM. http://www.eyesopen.com accessed April 2, 2013). (66) Tuccinardi, T.; Botta, M.; Giordano, A.; Martinelli, A. Protein kinases: Docking and homology modeling reliability. J. Chem. Inf. Model. 2010, 50 (8), 1432−1441. (67) ten Brink, T.; Exner, T. E. Influence of protonation, tautomeric, and stereoisomeric states on protein−ligand docking results. J. Chem. Inf. Model. 2009, 49 (6), 1535−1546. (68) Waszkowycz, B.; Clark, D. E.; Gancia, E. Outstanding challenges in protein−ligand docking and structure-based virtual screening. Wiley Interdiscip. Rev.: Comput. Mol. Sci. 2011, 1 (2), 229−259. (69) Totrov, M.; Abagyan, R. Flexible ligand docking to multiple receptor conformations: A practical alternative. Curr. Opin. Struct. Biol. 2008, 18 (2), 178−184. (70) Patel, R. Y.; Doerksen, R. J. Protein kinase-inhibitor database: Structural variability of and inhibitor interactions with the protein kinase P-loop. J. Proteome Res. 2010, 9 (9), 4433−4442. (71) McGann, M. FRED and HYBRID docking performance on standardized datasets. J. Comput.-Aided Mol. Des. 2012, 26 (8), 897− 906. (72) The PyMOL Molecular Graphics System, version 1.3; Schrödinger, LLC: New York. (73) VIDA, version 4.2.1; OpenEye Scientific Software: Santa Fe, NM. http://www.eyesopen.com accessed April 2, 2013). (74) Schrödinger Suite 2012: Maestro, version 9.3.5; Schrödinger, LLC: New York.

L


Docking Challenge: Protein Sampling and Molecular Docking

Recommend Documents